ABSTRACT
Contrary to the membrane-anchored leukemia inhibitory factor receptor (LIFR), the mouse soluble LIFR is an inhibitor of LIF action, possibly through a ligand titration effect. Two mRNA species encoding the soluble LIFR have been identified. Since the 3'-untranslated end of the shorter form was shown to contain a B2 element, we have examined the possibility that this SINE may be responsible for LIFR mRNA truncation. Transient expression assays, using B2-derived or intron-derived sequences independently or in conjunction, show that the B2 element has fortuitously unmasked a cryptic pre-mRNA 3' processing activity of silent intron sequences. The corresponding locus of the rat genome has been isolated and was shown to be devoid of any retroposon, which may explain why no soluble LIFR has yet been identified in any other species and further indicates that the B2 insertion event in the mouse LIFR gene has occurred recently during evolution. And yet, a tight tissue-specific regulation of alternative synthesis of soluble and membrane-bound LIFR mRNA has already emerged in mice. These results provide striking evidence for the rapid influence of retroposition on genome expression.
LIFR mediates the effects of several cytokines, like ciliary neurotrophic factor, leukemia inhibitory factor (LIF, D-factor), oncostatin M, interleukin 6 (1 ) and other unidentified factors (2 ). Accordingly, targeted disruption of the LIFR gene causes pleiotropic defects (2 ,3 ). LIF, the first identified ligand for LIFR, ensures various biological functions, which may sometimes appear as opposites, like induction of differentiation of certain leukemic cells (4 ,5 ) and, conversely, inhibition of differentiation of embryonic stem cells (6 ,7 ). LIF is particularly important during the early stages of embryonic development, since it inhibits the gastrulation of mouse embryos (8 ) but favors intra-uterine implantation of blastocysts (9 -11 ).
LIFR is a transmembrane cellular protein, converted into a high affinity receptor by association with the signal transducer gp130 (1 ). A soluble LIF binding substance (LBP) has been isolated from normal mouse serum. Possibly in relation with the influence of LIF on embryonic development, the concentration of LBP was shown to dramatically increase in pregnant mice (12 ), reaching a maximum of 20-30 times the basal level at 15 days of pregnancy (13 ). After biochemical purification, LBP turned out to correspond to a soluble, truncated form of cellular LIFR, devoid of the transmembrane and intracellular domains (12 ,14 ). This soluble LIFR was shown to inhibit LIF action by preventing its binding on target cells (12 ,13 ), raising the hypothesis that it may serve as an inhibitor of the systemic effects of locally produced LIF (12 ). Considering the high uterine production of LIF at this time (10 ), this hypothesis may explain the large increase in soluble LIFR concentration observed in serum during pregnancy.
Two mRNAs encoding soluble LIFR in mice have been isolated. One, of 2.6 kb, is truncated and polyadenylated further upstream than normal LIFR mRNAs. It contains in its 3'-untranslated region the sequence of a non-coding rodent-specific B2 retrotransposon (15 ,16 ). A second longer mRNA containing the B2 element has been identified, which carries additional 3' sequences encoding the transmembrane domain (17 ). This last observation strongly suggests that mRNAs coding for the membrane-anchored and the soluble types of LIFR derive from the same genome locus, through complex mechanisms.
In this report we have investigated the possible influence of the B2 element in the formation of mRNA isoforms encoding soluble LIFR.
Ten PCR primers were synthesized that correspond to exonic regions found in LIFR mRNA isoforms and whose positions relative to the LIFR gene are shown in Figures 1 and 2 (Us1, 5'-GGAGAAAGGTTCCTTCAAACAGCAC-3'; As1, 5'-GTTACGTTCCATAATTGGATACGTAGAAG-3'; Bs1, 5'-TATCCAGGCCTAACACCAGAG-3'; Bs2, 5'-CTTCTGGTGTGTCTGAAGAGAG-3';Bs3, 5'-CCTTCAGCCCTTGACTGCT- TG-3'; Ba1, 5'-TCTCTGGTGTTAGGCCTGGATAC-3'; Ba2, 5'-CACCATGTGGTTGCTGGGA-3'; Ba3, 5'-ACAAGCAGTCAAGGGCTGAAG-3'; Ca1, 5'-TCTTCCACGGGGATGTCGTC-3'; Da1, 5'-CTCAGTGTCTTCTGGGATATGTCAG-3'. Four genome fragments were obtained by PCR using the following primer pairs: Bs1/Ba3, Bs1/Da1, As1/Ba1 and As1/Ca1. Two cDNA fragments were obtained by RT-PCR using the primer pairs Us1/Da1 and Us1/Ba1. PCR products were inserted into the pGEM-T cloning vector (Promega Biotech) and sequenced in both orientations using the DyeDeoxy Terminator Cycle Sequencing kit Prism (Applied Biosystems) and an automated sequence apparatus (373; Applied Biosystems Inc.) with the PCR primers described above and plasmid primers SP6 and T7. Gel files collected by ABI 373A Data Collection Software were transferred and analyzed using GeneJockey II software (Biosoft, Cambridge, UK).
Genomic DNA was prepared as described (18 ) and RNAs using the TriZol reagent (Gibco BRL). RNAs were reverse transcribed into cDNA using random hexamers as primers and Superscript reverse transcriptase (Gibco BRL). RT-PCR primers were designed inside separate exons to avoid any bias due to residual genomic contamination. In particular, the 3'-most primer was chosen in exon D rather than exon C, since this last one is only juxtaposed to the alternative B exon in the genome. A poly(A) tail-specific primer dT [5'-(T)20A-3'] was also used to discriminate between the two LIFR mRNAs encoding soluble LIFR (types II and III).
The choramphenicol acetyltransferase gene was inserted into the XhoI site of the pCEP4 plasmid (Invitrogen Corp.), in an orientation allowing its transcriptional control by the CMV promotor. The SV40 transcription termination site was then deleted by digestion with BamHI and partial digestion with SalI and replaced by a SalI-BamHI fragment from the polylinker of plasmid pBSK (Stratagene), which reconstitute an intact BamHI site at this position. Then, various combinations of mouse LIFR gene fragments were inserted into this blunted BamHI site.
Mouse 3T3 cells were maintained in DMEM (Gibco BRL) supplemented with 5% newborn calf serum (Gibco BRL). They were co-transfected in 100 mm Petri dishes using the calcium phosphate-DNA co-precipitation procedure, with 3 [mu]g plasmids containing the CMV-CAT-LIFR constructs shown in Figure 4 and 2 [mu]g [beta]-actin-LacZ plasmid, used as an internal control of transfection efficiency. Twenty hours after transfection, cells were harvested for protein extraction. [beta]-Galactosidase activity and CAT protein content were then measured using respectively an in vitro assay (19 ) and the CAT ELISA kit from Boehringer.
Intron-exon organization of the LIFR gene region responsible for synthesis of the different types of LIFR mRNAs was determined from DNA sequence data comparisons between genome and cDNA fragments as described in Materials and Methods. The nucleotide sequence of this region with the major splice signals is displayed in Figure 1 . Comparison between the alternative regions of LIFR mRNAs proposed by Tomida et al. (17 ) and the corresponding genome region indicates that a DNA segment (named B in Fig. 1 ) is alternatively used as a part of an exon in form III of the transcript or as a part of an intron in mRNA form I. This B region is either maintained in mature mRNAs, fused to the C exon, or removed, together with the short intron following the A exon. A short intron (250 nt), located exactly at the 5'-end of the alternative B region, is present only in soluble LIFR-encoding mature mRNAs. The scheme presented in Figure 2 summarizes the different splicing events leading to the three possible transcripts (named I, II and III). This figure also shows the position of the RT-PCR primers and the size of the expected product used to monitor occurrence of these transcripts. The genome DNA sequence (Fig. 1 ) is compatible with all these splicing patterns, since two potential 3' splice sites (CAG) are present on both sides of the B region (named ss2 and ss1 in the figures). In this respect we have found an additional G residue at the 3'-end of the B region absent from the cDNA sequence reported by Tomida et al. (17 ). This change is functionally important, since it contributes to the alternative 3' splice site ss2. Finally, the last category of LIFR mRNA (II) described by Tomida et al. (17 ) indicates that the B region may also contain a site allowing premature 3' processing of the pre-mRNA.
The different assortment of exons in transcripts as described in Figure 2 should have profound consequences for the nature and structure of translated LIFR proteins. As predicted by Tomida et al. (17 ), LIFR mRNA type I encodes a cellular receptor, while types II and III encode soluble LIFR. The soluble LIFR form has a truncated transmembrane domain because of the presence of a stop codon, TAA, in-frame in exon B (Fig. 1 ). Specific detection of cellular and soluble LIFR-encoding mRNAs clearly shows that their accumulation is disconnected in different organs, suggesting that independent mechanisms of regulation apply to synthesis of the soluble and membrane forms of LIFR (Fig. 3 ). Soluble LIFR appears less widely distributed than cellular LIFR. Moreover, when present, soluble LIFR transcript may be either less abundant (in ovary, epididymis and retina), in similar amounts (in seminal vesicle and kidney) or strikingly more abundant than cellular LIFR (in liver). Remarkably, the ratio of soluble to cellular LIFR is inversely related in liver and testis. As shown in Figure 4 , liver contains predominantly soluble LIFR-encoding mRNAs, while in testis, where the couple LIF/LIFR plays a locally important role in spermatogenesis (20 ), no trace of soluble LIFR mRNA could be detected even with multiple rounds of RT-PCR and the Us1/Ba1 primer pair. Conversely, cellular LIFR transcript was detected in both organs with the Us1/Da1 primer pair. Furthermore, specific detection of soluble LIFR type II and III mRNAs, using respectively the As1/poly(dT) and Bs2/Ca1 primer pairs, reveals that the two forms are present in liver but absent in testis (Fig. 4 ), indicating that LIFR pre-mRNA premature 3' cleavage, as well as alternative splicing, do not occur in this organ.
We have looked for the presence of the different LIFR mRNA species in livers from different mouse strains and from rat. As shown in Figure 5 , the three forms can be found in all mouse strains tested, while only type I mRNA can be detected in rat. Such a complete absence of form II and III LIFR mRNAs in the liver from rat led us to check for the presence of a B2 element in the rat LIFR gene. To assess this point, we isolated, by genomic PCR with the As1 and Ca1 primers, the region of the rat LIFR gene homologous to that of mouse overlapping exons A-C (Fig. 1 ). The nucleotide sequence of rat and mouse PCR products was determined and aligned. As shown in Figure 6 A, the intron between exons A and C is globally well conserved between the two rodent genomes. Interestingly, the region corresponding to mouse alternative splice acceptor site ss2 is 100% identical in rat, with a CAG trinucleotide downstream of a T-rich DNA region reminiscent of the polypyrimidine stretch at the 3'-ends of introns.
When compared with other pre-mRNA 3' processing sites, the possible alternative cleavage region of the B exon giving rise to mRNA form II of LIFR has a bipartite organization, with an upstream AATAAA motif and a downstream G/T-rich domain. Moreover, the DNA sequence of these domains is highly analogous to that of a [beta]-globin gene ([beta]-maj.globin; Fig. 7 A) known to provide a strong 3' processing signal of the pre-mRNA (22 ). As shown in Figure 7 A, the cleavage sites are located at similar distances 3' of the consensus AATAAA (18 nt for the LIFR gene and 20 nt for [beta]-maj.globin) and both possess, downstream, similar G/T repeats. Thus, the polyadenylation site provided by insertion of the B2 element in the mouse genome might cooperate with downstream G/T-rich intron sequences to generate a new functional 3' processing site resulting in a truncated LIFR mRNA form that encodes the soluble LIFR. This hypothesis was investigated through pre-mRNA 3' cleavage functional assays.
We have presented evidence that the molecular mechanisms responsible for synthesis of the alternative forms of LIFR mRNA combine alternative 3' processing and alternative splicing. Synthesis of type III mRNA results only from alternative splicing, while that of type II mRNA requires in addition pre-mRNA cleavage at a site generated by cooperation between the B2 retroposon and surrounding intronic sequences. Moreover, we have shown that the mRNAs encoding soluble LIFR are synthesized in a tissue-specific manner, suggesting that activity of the B2-associated pre-mRNA 3' processing site is subject to stringent regulation.
Taken together, the above observations allow the proposal of a double involvement of B2 in diversified mouse LIFR mRNAs. A feature of LIFR pre-mRNA maturation into type II and III mRNAs is alternative 3' splice site selection, linking exon A to ss2 instead of ss1. When applied to the full-length LIFR pre-mRNA, this splicing event leads to synthesis of the type III mRNA. The absence of such molecules in rat (Fig. 5 ), although the rat intron does contain a ss2 site identical to that of mouse, strongly supports the idea that selection of this site in mouse is driven by the presence of a downstream B2 element (Fig. 8 , left panels).
Contrary to splicing enhancement, to our knowledge there is no previous example of a role of B2 elements in pre-mRNA cleavage. The results presented in this article are consistent with a model in which use of an optional exon (named B in Fig. 2 ) is controlled mainly by premature pre-mRNA cleavage at a chimeric site, made of a region from a B2 element and from intronic sequences. Pre-mRNA 3' processing sites are most often bipartite, so that the risk of inopportune pre-mRNA cleavage is prevented by a double key mechanism, where the simultaneous and synergistic binding of two factors, most often CPSF and CstF, binding respectively to AAUAAA and G/U-rich motifs, is necessary to trigger recruitment of cleavage factors (35 ). By themselves, B2 elements do not cause transcription termination, as demonstrated by their presence inside several primary transcripts, such as those of the rat growth hormone (36 ) or ornithine decarboxylase genes (37 ). In fact, B2 elements carry at their 3'-end multiple overlapping AATAAA motifs, often referred to as `polyadenylation sites', but which have been shown to be insufficient to induce transcription termination (38 ). The AATAAA box must be paired to either a convenient downstream G/T-rich motif or to an upstream sequence element (USE) (39 ) to acquire its pre-mRNA 3' processing activity. The inability of B2 elements to cause pre-mRNA cleavage indicates that the internal sequence of B2 does not contain any USE.
In the case of the mouse LIFR gene we have shown that insertion of B2 has occurred in the vicinity of a G/T-rich motif, identical to that found in the major termination site of the [beta]-globin gene, and is very efficient when compared with others (22 ). The exact nature of downstream GU-rich elements remains poorly defined. They most likely correspond to specific sequences rather than to mere regions with high GU content (38 ). This suggests that perfect alignment of a 9 bp motif between the GU elements from the [beta]-globin and mouse LIFR genes is not fortuitous. Thus, we hypothesize that random integration of a B2 element has revealed a cryptic downstream GU-rich motif in a LIFR gene intron convenient for pre-mRNA cleavage. This chimeric site, fully efficient in our transient expression assays, should, however, not be too strong in vivo to allow synthesis of the normal type of LIFR. This situation is reminiscent of the case of the upstream weak poly(A) signal of the alternatively polyadenylated immunoglobulin in the [mu] constant gene (40 ).
Although the B2 element should have integrated into the mouse LIFR gene relatively recently during evolution, as indicated by its absence in the rat locus, its has already acquired a biological function, allowing generation of a variant of LIFR endowed with alternative biological properties. Production of soluble LIFR is tightly regulated both in space, as between different organs like liver and testis, and in time, as in the liver of female mice depending on the pregnancy state (15 ). In addition to the pregnancy-dependent expression of soluble LIFR by liver (12 ,13 ), we have shown here that alternative use of the B2-associated poly(A) site is also tissue specific. This observation strongly suggests that pre-mRNA cleavage involves non-ubiquitous trans-acting factors, in addition to the basal components already identified (35 ). These findings also indicate that, as is the case for transcription promoters, the strength of pre-mRNA cleavage cis-elements is not an absolute parameter, but depends on the presence of tissue-specific or developmentally regulated factors. The case of the mouse LIFR gene provides a useful model for identifying cis- and trans-acting signals dictating tissue- and development stage-specific regulation of conditional pre-mRNA 3' processing.
Besides LIFR, many ligands and receptors are already known to co-exist as soluble and membrane-bound types, for example IL4 (41 ), NGF (42 ) receptors and the c-kit ligand (43 ). Various strategies have been elaborated to allow synthesis of these different types from the same genome loci, including elimination of the transmembrane-coding domain by alternative splicing, proteolytic cleavage of the extracellular domain or a combination of these two mechanisms, as in the case of the c-kit ligand, where alternative splicing allows retention or exclusion of an exon coding for a short extracellular domain and containing a specific proteolysis site (44 ). The case of soluble LIFR provides another, unexpected mechanism by which to generate a soluble receptor and demonstrates that random transposition mechanisms can rapidly create highly regulated, novel biological functions. In particular, the dual ability of B2 elements to enhance splicing at an upstream site and to create pre-mRNA cleavage sites makes them highly liable to generate new possibilities for gene regulation by insertion inside introns.
REFERENCES